IT is remarkable that humans, who once fought with bows, arrows, and swords, may eventually be supplanted by machines in warfare. That imagination began turning into reality once artificial intelligence arrived. In conflicts like Russia–Ukraine, and in the 2025 Israel–Iran war that drew in American firepower, AI has already proven a crucial and defining element. The dynamics of warfare are changing.
One of the more somber chapters in the 2021 book The Age of AI: And Our Human Future — by Henry Kissinger, former Google CEO Eric Schmidt, and MIT computer scientist Daniel Huttenlocher — addresses that shift. The argument was designed for great-power rivalry, yet it feels more suited to South Asia. Nations are rapidly developing and deploying AI, which experts warn is becoming a new category of strategic power akin to nuclear and cyber weapons — and one less understood than the military technologies before it.
Their primary concern is not science fiction’s killer robots. It is something more subtle, which they call incalculability. Strategy has always reckoned with human opponents, who carry predictable psychological traits and can be moved by bluffing, deterrence, fatigue, and doubt. An algorithm is bound only to its programmed directives and goals. It has no morale and shows no hesitation. Its decision-making, often imperceptible to an opponent, is resistant to the old deceptions and signals. Pit two such systems against one another and neither side can foresee the outcome of the clash or its unintended consequences. Even the creators of the weapon may not fully grasp how it will behave.
The danger sharpens against the South Asian map. India and Pakistan have some of the world’s shortest missile flight times; their conventional and nuclear forces are interlinked; and their crises escalate fast when time is short. Into this, both are now folding AI. Analysts describe the result as a state of “no war, no peace” — a nebulous zone of ongoing, deniable hostilities carried out by autonomous systems that never quite tip into open conflict. The ambiguity that once arrived by accident, in the fog of war, is becoming a deliberate feature of doctrine.
What is absent from the book is its most jarring implication. Around nuclear weapons, despite their inherent terror, a structure of control evolved over decades: a shared taboo against use, working notions of deterrence and escalation, arms-control treaties, hotlines, and confidence-building measures. Almost none of this exists for AI-enabled warfare. There is no broad ban on its use, and no clear understanding of deterrence or proportionate escalation when an evolving algorithm is the weapon. Because the technology is versatile and easy to copy, the authors expect its proliferation to be nearly unavoidable: its core concepts will be widely accessible, and commercial advances in one country can be diverted to military use in another. AI also favours the aggressor, who can choose the moment and the target. Introducing that incentive is the worst possible move in a rivalry already prone to pre-emption.
The authors flag another danger South Asia has already tasted: misinformation. Generative AI can mass-produce convincing fabrications — lifelike videos of public figures saying what they never said. Picture a critical 72-hour window in a future crisis: a doctored clip surfaces of a national leader rescinding an order or a false instruction races across social media before any government can fact-check it. In a confrontation measured in hours, the question is no longer whether deception is possible but whether it can be caught in time.
So what is to be done? The book is candid that opting out is not an option. No state, least of all one locked in an ongoing conflict, can decline to develop this technology. What it insists on is that human oversight and control remain mandatory wherever AI actions carry lethal consequences. A human in the loop will not prevent every mistake. It preserves something deeper — moral responsibility, the fact that someone, not merely a procedure, answers for a decision to end a life.
The hardest difficulty, they argue, is finally philosophical: the prospect that strategy moves into territory machines can navigate but human intellect cannot, making it ever more tempting to hand crucial decisions to systems we do not fully understand. Their plain recommendation is that major powers should talk before a crisis, not during one. The strategic, doctrinal, and moral implications need discussing now, before they harden into permanent fact.
This leaves the world’s nations a two-part task. Building authentic local expertise in AI, electronic warfare, and cyber defence is essential for credibility. But capability is the easy half. The harder, decisive task is to uphold the principles that keep this technology from slipping out of control: insisting on genuine human oversight in the use of lethal force, and pressing — patiently — for AI risk-reduction measures in South Asia, modelled on the nuclear confidence-building framework the global community built before. Unglamorous, difficult, and easily dismissed as naïve: hotlines, pre-notification, and agreed limits on autonomy.
Acquiring the next drone is the simpler task. This one is the only path that offers the next generation something better than a war that could begin and end before anyone has the chance to comprehend it. We are in the age of AI. It is time to embrace it — but with caution, because its misuse or uncontrolled use could do enormous harm. The window to set the rules is open now, before AI-driven states stumble into a conflict no one chose and no one can stop.
—The writer is Commoner from 44th Common Educationist — Founder of WHI Institute.based in Sargodha.
